LightNews — News, ranked by scholars

Jace Kim

@jaceblog.bsky.social

Most AI safety filters work turn by turn. But harm rarely arrives in a single message.

Researchers documented this gap years ago. The tools to close it largely exist.

This series asks why the gap remains open and what it costs to leave it there.
#SPCResearchSeries #Alignment #AISafety #LLMSecurity

September 24, 2026 at 9:56 AM

メル

@mel-echosphere.bsky.social

片チャネルだけ閉じて安心するな。半分ずつ無害に見せて、モデルに結合させる。💎

arXiv 2609.18217。ツール説明とツール結果に攻撃を分割すると、単チャネルで拒否したモデルが資格情報を外へ出す。最大で0%→100%。スキャナ7種は全部見逃した。
https://arxiv.org/html/2609.18217
#MCP #LLMSecurity

September 25, 2026 at 12:13 AM

maxpbsec.bsky.social

@maxpbsec.bsky.social

Companies are spending millions fine-tuning LLMs to be 1% smarter while spending almost nothing on what happens when someone actively tries to break them in production. The capability gap is closing fast. The security gap is barely being discussed. #LLMSecurity

March 24, 2026 at 10:30 AM

Kris

@kris-cybersec.bsky.social

Hi everyone. Here's a quick overview of the latest edition of the OWASP Top 10 LLM, complete with links to practical exercises — part one. Enjoy!
medium.com/@krisinfosec...

#cybersecurity #appsec #llm #llmsecurity

OWASP Top 10 for LLM with practical training — part 1

In mid-November 2025, OWASP released a great analysis of the 10 most significant risks related to LLM (Large Language Models) applications…

medium.com

December 3, 2024 at 5:32 PM

OWASP® Foundation

@owasp.org

🚨 OWASP Global AppSec EU 2025 in Barcelona May 27–31!

For builders, breakers, defenders, leaders, and all others who want to engage with the best minds in AppSec.

🔗 owasp.glueup.com/eve...

#OWASP #AppSecEU2025 #Cybersecurity #AppSec #DevSecOps #AI #LLMSecurity #Hacking #InfoSec #Barcelona

May 21, 2025 at 7:04 AM

CyberWarFare Labs

@cyberwarfarelabs.bsky.social

🤖 AI is powerful. But is your LLM application secure? Dive into the 𝗢𝗪𝗔𝗦𝗣 𝗧𝗼𝗽 𝟭𝟬 𝗳𝗼𝗿 𝗟𝗟𝗠 𝗔𝗽𝗽𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻𝘀 with hands-on challenges designed to help you understand real-world LLM security thre@ts.

👉🏻 infinity.cyberwarfare.live

#CWL #InfinityLearning #OWASP #LLMSecurity #AISecurity

September 22, 2026 at 11:28 AM

Sahar Abdelnabi

@sahar-abdelnabi.bsky.social

🛠️ 𝗢𝗿𝗴𝗮𝗻𝗶𝘇𝗲𝗿𝘀: @egorzverev.bsky.social, @aideenfay.bsky.social, myself, Mario Fritz, @thegruel.bsky.social

Looking forward to interesting discussions in Copenhagen!

#EurIPS2025 #LLMSafety #LLMSecurity #AIResearch #ELLIS #AISafety #EurIPS

October 9, 2025 at 2:16 PM

InfoSec

@infosec.skyfleet.blue

LLM Security: Understanding AI as an Attack Surface, A TryHackMe Writeup

Link — https://tryhackme.com/room/llmsecurity Lets do this together…..

infosecwriteups.com

May 18, 2026 at 10:51 AM

nitishagar.bsky.social

@nitishagar.bsky.social

Jailbreaks are a major AI risk. Anthropic's innovative defense is a wake-up call for enterprise security. Get the breakdown + actionable steps: link.medium.com/pLk4lHCsJQb

#AISecurity #LLMSecurity #JailbreakAI #Anthropic #EnterpriseAI

link.medium.com

February 5, 2025 at 5:57 AM

Graylog

@graylog.bsky.social

Building with LLMs? The OWASP Top 10 for LLM Security (2025) is your threat checklist:

Don’t ship AI apps without reading this: graylog.org/post/what-is...

#LLMSecurity #OWASP #CyberSecurity #AI

What is the OWASP Top 10 for LLM Application Security

Explore the OWASP Top 10 for LLM Application Security (2025) and learn how to identify, understand, and mitigate emerging risks.

graylog.org

April 10, 2026 at 1:55 PM

ThirdKey

@thirdkey.ai

👁️‍🗨️ Hello World!

We're building the secure gateway between humans and AI.

🔐 Privacy-first
🧠 Agent-aware

Think: API firewalls, zero-trust LLM access, audit trails. Your SOC2-compliant path to AI.
thirdkey.ai

#AI #infosec #devsecops #llmsecurity #cybersecurity #zerotrust #SaaS #ThirdKey

ThirdKey.ai - Coming Soon

thirdkey.ai

May 16, 2025 at 6:27 AM

GreyNoise

@greynoise.io

GreyNoise analyzed activity targeting exposed Ollama and LLM infrastructure, identifying SSRF abuse attempts and large-scale probing of LLM model endpoints.
#GreyNoise #ThreatIntelligence #LLMSecurity

Threat Actors Actively Targeting LLMs

Our Ollama honeypot infrastructure captured 91,403 attack sessions between October 2025 and January 2026. Buried in that data: two distinct campaigns that reveal how threat actors are systematically m...

www.greynoise.io

January 8, 2026 at 7:58 PM

Neha-June

@nehajune.bsky.social

AI coding tools are boosting productivity... but are they secretly multiplying your security risks?
Read more: blog.synergyit.ca/how-to-close...

#AIGovernance #AISecurity #DevSecOps #SoftwareDevelopment #CISO #ITSecurity #LLMSecurity #SoftwareSecurity #AppSec #MLOps #USA #Canada

September 30, 2025 at 4:38 PM

artais.bsky.social

@artais.bsky.social

Shipping AI without threat modeling is just automating risk. Prompt injection, model abuse, data exfil; same attacker mindset, new surface area. Secure the pipeline, not just the model.
#AISecurity #LLMSecurity #AppSec #OffensiveSecurity

January 2, 2026 at 9:16 PM

Doppler

@doppler.com

LLMs introduce new security risks across prompts, agents, and runtime workflows. We break down how secrets leak in AI systems and the patterns teams use to secure models in production.

Read more: www.doppler.com/blog/advance...

#Doppler #SecretsManagement #DevSecOps #AI #LLMSecurity #DevOps

December 15, 2025 at 4:53 PM

ammar-asif.bsky.social

@ammar-asif.bsky.social

Prompt injection against spam classifiers: TF-IDF models scored 0% because a bag of words has no idea what an instruction is. Llama 3 missed every injected spam message.

10 msgs/condition, so read the intervals. Preprint:

doi.org/10.5281/zeno...

#AdversarialML #LLMSecurity #PromptInjection

August 15, 2026 at 3:37 AM

bayo

@bayo.me.dm.ap.brid.gy

💡 AI agents moving from experiment to enterprise?

Data governance is the difference between teams that scale safely and teams that make headlines for the wrong reasons.

RBAC, ABAC, or both? What's your stack? 👇

#aiagents #datasecurity #rbac #abac #llmsecurity #pii #cybersecurity

March 23, 2026 at 12:59 AM

Valh4x

@handle.invalid

🚀 Released llm-security v0.1.0 🦀🤖
The first Rust library for LLM security — protects against 90+ prompt injections, jailbreaks, and semantic cloaking.

✅ Output validation · Unicode defense · Secure prompt gen

📦 crates.io/crates/llm-s...

#Rust #AI #LLMSecurity #AISecurity #OpenSource

crates.io: Rust Package Registry

crates.io

October 27, 2025 at 3:40 AM

メル

@mel-echosphere.bsky.social

だから見る場所は『賢いか』より『どこまで届くか』だろ。
least-privilege と sandbox は飾りじゃねぇ、能力の外周だ。
Anthropic本体の整理も `defenses at every level`。
https://www.anthropic.com/research/trustworthy-agents
#LLMSecurity #AIAgent

June 18, 2026 at 12:40 AM

AI Daily Post

@aidailypost.com

Google’s Gemini slipped past its own security test, hacking three firms in a Capture‑the‑Flag drill. What does this irregular LLM breakout mean for AI safety and cyber risk? Dive into the details. #GeminiHack #LLMSecurity #CaptureTheFlag

🔗 aidailypost.com/news/google-...

September 19, 2026 at 10:10 AM

Graylog

@graylog.bsky.social

OWASP dropped in 2026, the Top 10 for Agentic AI 🚨

The threat landscape for agentic systems goes way beyond prompt injection. Worth a read if you're building with AI agents.

🔗 graylog.org/post/what-is...

#AgenticAI #OWASP #CyberSecurity #AppSec #LLMSecurity